Observation: Summary

Naturalistic observation refers to collecting data without interfering with the ongoing behavior. Observation relies on information available to the senses -- sight, hearing, touch, etc. It provides information about the "what" of behavior, but does not reveal it underlying motivation (the "why").

Observational research begins with casual observation - checking out the setting and the activities that occur there. To collect valid data, it is important to distinguish between observation and inference. Systematic observation uses decision rules and operational definitions in order to reduce error. Reliability of observational data is estabilshed by having at least two observers independently record data and then checking for a high level of agreement (termed interobserver reliabiity).

Two general categories of potential error in observation are bias (from the observer) and reactivity (on the part of the observed). Observer bias is reduced by making clear rules about procedure, having a well-defined checklist of behaviors, setting a time frame, and training observers. Reactivity on the part of those being observed is reduced by the observer's being unobtrusive or relying on habituation so as to be no longer noticed.

Review Methods Manual: Guidelines to systematic observation

Participant observation is different from systematic observation. The observer participates in the activities and events being observed.Because of the level of involvement of the observer, it is particularly subject to problems of bias and reactivity. The data generally consist of an individual's notes, and thus may not be as reliable as information gathered by two or more independent observers. Participant observation provides descriptive qualitative data (in contrast with the numeric information often gained by systematic observation). Ethnography, the study of particular people and places, often uses participant observation as a research method. Quantitative (numerical) and qualtiative (narrative) approaches are different, but are not mutually-exclusive. Both methods can provide useful information about behavior.

Terms to know (define each before clicking to see the definition in a pop-up message)